CO2 Stripper Postmortem Thoughts

[EDIT: A cru­cial con­sid­er­a­tion was pointed out in the com­ments. For all the de­signs I’ve looked at, it’s cheaper to just get a heat ex­changer and ven­tila­tion fans, and blow the air out­side/​pull it in­side and eat the ex­tra heat­ing costs/​throw on an ex­tra layer of cloth­ing, than it is to buy a CO2 strip­per. There’s still an ap­pli­ca­tion niche for poorly ven­tilated rooms with­out win­dows, but that de­scribes a lot fewer oc­ca­sions than my pre­vi­ous dreams of com­mer­cial use.]


So, I have fi­nally com­pleted build­ing a CO2 strip­per that re­moves CO2 from the air to (hope­fully) im­prove cog­ni­tion in en­vi­ron­ments with high CO2 lev­els. In Cal­ifor­nia, the weather is pretty good so it’s easy to just crack a win­dow at any point dur­ing the year, but other ar­eas get quite cold dur­ing the win­ter or quite warm dur­ing sum­mer and it’s in­fea­si­ble to open a win­dow un­less you want to spend an awful lot of money on heat­ing or cool­ing bills. It didn’t work quite as well as the math in­di­cated at first, but the whole thing is built, and ba­si­cally func­tional. The rest of this post will be a re­flec­tion on the les­sons learned while do­ing so.


1: In hard­ware, ideas are cheap, ex­e­cu­tion is expensive

So, the fun­da­men­tal idea is ex­tremely sim­ple once you have some ba­sic knowl­edge of chem­istry. The goal is to get CO2 into some form that isn’t the gas form, via some sort of chem­i­cal re­ac­tion.

Sub­marines and CO2 cap­ture from flue gas use a re­versible re­ac­tion with ethanolamines, where they ab­sorb CO2 at high tem­per­a­tures and re­lease it at low tem­per­a­tures. Rev­ersible re­ac­tions are good for mak­ing waste, but heat­ing up and cool­ing down large quan­tities of liquid takes an awful lot of en­ergy. Sub­marines have nu­clear re­ac­tors on­board, and flue gas is hot, but we don’t nec­es­sar­ily have the en­ergy re­quired. Also ethanolamines are toxic and hard to get a hold of for a civilian and re­ally stinky, be­ing the ma­jor com­po­nent of “sub­marine smell”.

Ad­sorp­tion onto ze­o­lites is also plau­si­ble, but the is­sue is that it re­quires al­ter­nately ex­pos­ing the ze­o­lites to high air pres­sure and low air pres­sure, and high airflow is re­quired. The com­bi­na­tion of high pres­sure and high airflow means that again, you’re us­ing a lot of en­ergy. The ba­sic math is as fol­lows: One hu­man pro­duces about 1 kg of CO2 in 24 hours. We can ideal­ize a perfect CO2 strip­per as a magic box that in­hales air and spits it out at 0 ppm. If you want a steady-state con­cen­tra­tion of 500 ppm for 2 peo­ple, then we can see how much air-flow is re­quired to lock up 2 kg of CO2 in 24 hours. This comes out to about 100 cu­bic feet per minute. This is the bare min­i­mum air flow for any CO2 strip­per, but in this par­tic­u­lar case, it cor­re­sponds to a 25 horse­power air com­pres­sor, which is 18 kilo­watts. This is equiv­a­lent to run­ning 5 elec­tric dry­ers at once. So that one is out too, es­pe­cially since we were as­sum­ing 100% effi­ciency at elimi­nat­ing CO2.

What about ir­re­versible re­ac­tions? Just lock the CO2 up as a solid waste? Well, to be­gin with, this is go­ing to pro­duce quite a waste stream, and con­sume quite a bit of chem­i­cals, you’d bet­ter hope it’s safe and that the feed chem­i­cal is cheap. The re­ac­tion used on space mis­sions used lithium hy­drox­ide. The ba­sic idea is that lithium hy­drox­ide makes a very ba­sic solu­tion. Car­bon diox­ide is slightly acidic, so it dis­solves very fast into ba­sic solu­tions. Then you get pre­cip­i­ta­tion of lithium car­bon­ate which is safe.

The prob­lem is that lithium hy­drox­ide is quite ex­pen­sive. It was used on space mis­sions be­cause it’s the most mass-effi­cient way of do­ing that sort of re­ac­tion and ev­ery gram counts in space mis­sions, but we want the cheap­est way of do­ing that re­ac­tion.

And then we hit upon the perfect solu­tion. Cal­cium hy­drox­ide. It’s an ex­tremely cheap bulk chem­i­cal, 15 bucks for a 50-pound sack of it at a hard­ware store. It’s fairly mild as far as hy­drox­ides go, be­ing pH 12.4. So in­stead of giv­ing you hor­rible chem­i­cal burns, it’s safe to han­dle un­less you’re ex­posed to it for over an hour at a time with­out wash­ing it off. It’s the alkaline analogue of the differ­ence be­tween 1 M hy­drochlo­ric acid, and lemon juice. And when it re­acts with CO2, it makes CaCO3, aka lime­stone, which is to­tally harm­less. In fact, it’s a com­mon lab­o­ra­tory demon­stra­tion that breath­ing onto a solu­tion of this stuff pro­duces a white film/​crust on the top, which is the CO2 in the breath locked up as solid lime­stone. It’s the ob­vi­ous choice if you’re try­ing to re­move CO2 via chem­i­cal means.

And in fact, in the SSC com­ment sec­tion, some­one else in­de­pen­dently had the ex­act same idea! Just lock up CO2 with cal­cium hy­drox­ide!

The sim­plic­ity of an idea in the field of atoms in­stead of bits doesn’t nec­es­sar­ily mean that any­one on earth has ever done it be­fore, though, or will ever do it, and I’m not wor­ried about any­one scoop­ing the idea, be­cause build­ing novel hard­ware is hard enough to provide a nat­u­ral bar­rier to en­try un­less it’s a large com­pany that’s in­ter­ested in the idea. Ideas are cheap, ex­e­cu­tion is ex­pen­sive, in both time and money.

2: Only poly­maths need apply

If you’re try­ing to build a novel ma­chine in your garage, and aren’t work­ing as part of an en­g­ineer­ing team, you will ei­ther need an im­prob­a­bly wide range of knowl­edge, or the gen­eral abil­ity to pick up what­ever you need to learn. There’s the ba­sic knowl­edge of chem­istry to spot that this is the ob­vi­ous re­ac­tion to go for, but the full de­sign re­quires:

Fa­mil­iar­ity with wastew­a­ter aer­a­tors to know what to buy to pre­vent clog­ging with solids, knowl­edge on which ma­te­ri­als won’t re­act with your chem­i­cals, the math of air flow in pipes, the abil­ity to read fan pres­sure/​airflow curves, the abil­ity to go from “I want a cir­cuit that does this” to build­ing a novel elec­tronic cir­cuit on a bread­board with­out fry­ing any­thing im­por­tant, enough pro­gram­ming knowl­edge to write some ba­sic ar­duino code, fa­mil­iar­ity with haz­ardous waste dis­posal reg­u­la­tions in your state, fa­mil­iar­ity with waste de­wa­ter­ing tech­niques, fa­mil­iar­ity with which sort of pumps can pump sludge in­stead of pure wa­ter, some elec­tri­cal en­g­ineer­ing knowl­edge to work safely with 220V power with­out fry­ing your­self or any­one else, knowl­edge of sound­proofing, and es­pe­cially the fa­mil­iar­ity with ev­ery­thing at Home De­pot that lets you home in on the most effi­cient and foolproof way of build­ing a thing that does what you want. Prob­a­bly some other stuff too that I con­sider ob­vi­ous but oth­ers might not.

Now, most of this is pretty easy to pick up given enough start­ing men­tal fire­power, and the sense of what to google for. Or just hav­ing lots of ex­pe­rience with build­ing ma­te­rial things.

Hav­ing one of the rele­vant fields of knowl­edge man­i­fests it­self as know­ing ahead of time which ap­proaches will work and which will fail and what solu­tions past work in the area has already found.

For some of these, miss­ing it will man­i­fest as not know­ing that there’s an in­com­ing bul­let in a par­tic­u­lar area, like not know­ing that fine bub­ble aer­a­tors will promptly clog if there’s lots of par­tic­u­lates in the wa­ter, or not sus­pect­ing that high air flow rates are in­com­pat­i­ble with small pipe (I knew the lat­ter one and it still al­most got me un­til I idly de­cided to work out airflow ve­loc­ity in the pipe and re­al­ized it was around 200 mph)

3: The plan­ning fal­lacy is huge here.

So, it wound up cost­ing a lot more than I thought and tak­ing a lot longer than I thought. The mechanism of why the plan­ning fal­lacy hits so hard here is tied in with the de­sign pro­cess. What hap­pens is that you start out with a sketchy out­line of all the com­po­nent parts (like, “I need some­thing that au­to­mat­i­cally dis­penses chem­i­cal pow­der”), and as it be­comes time to build a part, you drill down fur­ther and fur­ther in flesh­ing out the de­tails un­til even­tu­ally you’ve drilled down far enough for your de­sign to Ac­tu­ally Work in re­al­ity. While you do this, you will in­evitably come across parts that are a lot harder to do than you ex­pected, which you were gloss­ing over on the first pass. The shiny black box of “build a chem­i­cal dis­penser” looks more tractable than “how the fuck do I build a mo­tor mount­ing plate with my in­ad­e­quate tools”, which you didn’t ini­tially sus­pect you had to do be­cause you weren’t think­ing at that level of de­tail. And also as you ad­dress the parts that are easy to do, all that is left is the parts that are hard or an­noy­ing or time-con­sum­ing to do, which can be some­what de­mor­al­iz­ing.

Same sort of thing goes with cost. You start out with “so here’s the cost for the big parts and ev­ery­thing else that’s left shouldn’t cost that much” (black-box warn­ing on “ev­ery­thing else”!), and then you go to Home De­pot and pick up a bunch of 4-inch ABS pipe and black glue and all the 90 de­gree and T pieces you need for the aer­a­tion pipes and look at the cost and it’s 100 bucks. Home De­pot trips add up shock­ingly fast. There’s also all the stuff you buy that you don’t even­tu­ally end up us­ing be­cause the de­sign evolves as you ac­tu­ally try to build it, like buy­ing gears when you don’t ac­tu­ally need gears, and all the stuff you didn’t think you had to buy but it turns out that you do need it.

And some­times you just get hit with some prob­lem you didn’t ex­pect at all and now have to fix, like “my fan is mak­ing a scream­ing noise, what do”

4. Why is there a valley of death?

Univer­si­ties and the gov­ern­ment funds ba­sic re­search. Then there’s the pri­vate sec­tor of busi­ness. The gap be­tween the two, where you have to go from ba­sic re­search to a busi­ness sel­l­ing the new ex­cit­ing thing is called the “valley of death”. Now, you’d think this is what R&D is for. But a lot of R&D from a busi­ness seems to be fo­cused on marginal im­prove­ments to ex­ist­ing things that already fall un­der the scope of what the ex­ist­ing busi­ness does, and not so much on build­ing a novel thing that can be the seed of a new busi­ness. Build­ing a novel thing re­quires a wide knowl­edge base, as dis­cussed be­fore, and in­evitably takes a lot more money and time than ex­pected. It’s the sort of thing done by in­ven­tors in a garage as a pro­ject of love, not the sort of thing you get paid to do.

Fur­ther, cross­ing the valley of death re­quires both the tech­ni­cal ca­pac­ity to build the thing, and the busi­ness skills to make a new busi­ness from scratch. If you have sev­eral peo­ple with differ­ent skills joined to­gether, it can be bridged, but one flaw of do­ing it alone is that there are a lot more in­ven­tors with the abil­ity to build the thing, than in­ven­tors with the abil­ity to build the thing and also the abil­ity or will­ing­ness to start a busi­ness that sells the thing. I’m in the former cat­e­gory. I can build it, but I hate build­ing it and if I have to build all the ma­chines my­self to sell, I’d flatly re­ject it, and I re­ally don’t want to be re­spon­si­ble for run­ning a busi­ness sel­l­ing it, I’d have no idea how to run a busi­ness, and it’d eat too much time. My dream is to get a de­sign good enough to sell, patent it, find some­one will­ing to make a busi­ness out of it, and just re­ceive a cut of prof­its with­out hav­ing to be in­volved in any­thing more re­gard­ing the pro­duc­tion or sel­l­ing of the ma­chines, be­sides helping out with tech­ni­cal de­sign work. Fur­ther, some­one with just the busi­ness skills won’t nec­es­sar­ily have the tech­ni­cal abil­ity to come up with the ma­chine in the first place, let alone build it. And there’s also the lemon mar­ket prob­lem of busi­ness­peo­ple iden­ti­fy­ing com­pe­tent non-scam tech­ni­cal peo­ple with a vi­able de­sign, and tech­ni­cal peo­ple find­ing com­pe­tent non-scam busi­ness­peo­ple.

There are fur­ther is­sues such as de­sign­ing the new in­ven­tion such that it is ro­bust and keeps work­ing for a while (not a prop­erty that pro­to­types gen­er­ally have), and de­sign­ing it such that it is easy to build and main­tain (also not usu­ally a prop­erty as­so­ci­ated with garage pro­to­types).

I’ve heard that there’s a com­pany in the UK that takes garage pro­to­types and up­dates the de­sign for ro­bust­ness, easy con­structibil­ity, and cost, which seems like an im­por­tant part of clos­ing the valley.

5. Build­ing alone vs build­ing as part of a team.

In a cer­tain sense, I was blessed on this pro­ject, be­cause I had com­plete con­trol over the en­tire de­sign. I had to con­tend with no meet­ings, and no un­ex­pected changes to parts of the de­sign that were already locked in, and no team de­ci­sions that were dumb and couldn’t pos­si­bly work. It’s the dream for any­one who dis­likes group pro­jects in en­g­ineer­ing. All failures are at­tributable to me alone, as well as all suc­cesses. Then again, hav­ing some­one else to work on the pro­ject with me definitely would have sped it up and I could rely on their knowl­edge of things I was ig­no­rant of, re­lax­ing the poly­math re­quire­ment. Maybe there’s an op­ti­mal de­sign team size? I guess it’d de­pend on how par­alleliz­able the work is, as well as how de­ci­sion-mak­ing-qual­ity scales with group size.

6. Fi­nal di­ag­no­sis and where to go from here.

So, it was over-time and over-bud­get and didn’t work as well as I had hoped, but it does in­deed work. Plan­ning fal­lacy is a huge ob­sta­cle here, and I now cer­tainly see why there’s a valley of death for this sort of work.

In or­der to make a ver­sion that’s prac­ti­cal for do­mes­tic use, I’d have to redo the de­sign to be a rain-column de­sign, pri­mar­ily be­cause it only re­quires high airflow, in­stead of the com­bi­na­tion of high airflow and high pres­sure, which re­quires buy­ing an ex­pen­sive fan from China and the ex­pen­sive elec­tronic com­po­nents which provide the ap­pro­pri­ate power to op­er­ate the fan. A rain column de­sign could use a much cheaper and sim­pler fan that op­er­ates from a wall out­let.

Fur­ther, in or­der for oth­ers in­ter­ested in CO2 re­duc­tion to have one of their own, I’d have to team up with some­one who could make a small busi­ness in as­sem­bling and sel­l­ing these things, prefer­ably in­volv­ing some­one who is not me build­ing the rele­vant thing. PM me if in­ter­ested.